Open
Conversation
…hen inappropriate.
AJBats
added a commit
to AJBats/saturncc
that referenced
this pull request
Apr 18, 2026
Three compiler changes that close the remaining gap between our
output and Hitachi SHC's, plus the first byte-identical corpus
match in project history.
1. sh_elim_redundant_ext: extend the sign-extension tracker to
recognize .w and .b displacement-mode (`@(disp,rN)`) and
indexed-mode (`@(r0,rN)`) loads as sources. Previously only the
indirect (`@rN`), GBR-relative, and pool-load forms were
recognized; .w disp/indexed loads from Gap 0/Gap 2 work were
leaking redundant exts.w through the pass.
2. sh_fuse_mov_into_add: new peephole. Collapses `mov rA,rB;
add rB,rC` into `add rA,rC` when rB is dead after the add
(forward liveness scan, bails at labels/branches). Common
shape when a .w/.b load's r0 result is copied to a temp and
then added to an accumulator — the copy is pure waste.
3. sh_elim_dead_labels: new peephole. Deletes label-only lines
(sh_is_label_line == 1) whose label is never referenced by any
other line. Token-boundary reference check via strstr with
alphanumeric/underscore lookahead (so `L1` in `L10` isn't a
false positive). LCC emits a function-exit label before the
epilogue for multi-return functions; single-return functions
leave it as dead code.
Impact vs prod (tier-1 byte-match):
FUN_06044834: 20 -> 10 -> 0 <-- byte-identical
FUN_00280710: 9 -> 9 -> 6
FUN_06000AF8: 17 -> 17 -> 16
FUN_06047748: 28 -> 28 -> 26
FUN_0604025C: 30 -> 30 -> 29
FUN_0602A664: 139 ->139 ->130
FUN_06040EA0: 261 ->261 ->260
FUN_06044BCC: 459 ->445 ->438
FUN_06037E28:1074 ->1062->1053
9 of 10 corpus functions improved, 0 regressed. Broad corpus
smoke 168/168 pass (unchanged). All 12 stage-4 regression tests
pass. Subagent independently graded all three sampled functions
IMPROVED with no semantic hazards.
FUN_06044834 is the project's first byte-identical corpus match —
closes the proof-of-thesis milestone tracked in
methodology_remediation.md.
Our output matches prod's 10 instructions exactly:
mov.w @(14,r4),r0 mov r0,r1
mov drh#26,r0 mov.w @(r0,r4),r0 add r0,r1
mov drh#30,r0 mov.w @(r0,r4),r0 add r0,r1
rts neg r1,r0
The path: Gap 0 (int→char* refactor, 385eafb), Gap 2 (#pragma
sh_weird_rule_1, b67fe7e), Gap 9 extension + mov-fuse + dead-label
elim (this commit). Four separate layers — C source, custom
pragma, peephole pass, sign-extension tracker — composed cleanly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This patch some updates from the plan 9 cpp back ported to the lcc cpp.